Inferring species abundance distribution across spatial scales

A long-standing problem in ecology is to understand how the species-abundance distribution (SAD) varies with sampling scale. The problem was first characterized by Preston as the veil line problem. Although theoretical and empirical studies have now shown the nonexistence of the veil line, this problem has generated much interest in scaling biodiversity patterns. However, research on scaling SAD has so far exclusively focused on the relationship between the SAD in a smaller sampling area and a known SAD assumed for a larger area. An unsolved challenge is how one may predict species-abundance distribution in a large area from that of a smaller area. Although upscaling biodiversity patterns (e.g. species-area curve) is a major focus of macroecological research, upscaling of SAD across scale is, with few exceptions, ignored in the literature. Methods that directly predict SAD in a larger area from that of a smaller area have just started being developed. Here we propose a Bayesian method that directly answers this question. Examples using empirical data from tropical forests of Malaysia and Panama are employed to demonstrate the use of the method and to examine its performance with increasing sampling area. The results indicate that only 10-15% of the total census area is needed to adequately predict species abundance distribution of a region. In addition to species abundance distributions, the method also predicts well the regional species richness.